Model details
GPT-6 Luna
GPT-6 Luna is a lower-tier member of OpenAI's GPT-6 family, released on September 22, 2026 alongside GPT-6 Sol as part of an expanded lineup that sits beneath the flagship GPT-6 Astra. Reporting describes Luna as a mid-priced, cost- and speed-optimized variant sharing the same underlying training approach as the flagship, but tuned for common enterprise tasks that demand less raw capability than Astra delivers. This positions Luna as a pragmatic choice for teams that want GPT-6-era improvements without paying flagship rates.
The defining advantage of GPT-6 Luna is economic efficiency paired with measurable capability gains. OpenAI cut API pricing to roughly half or less of the GPT-5.6 generation, made possible by inference and caching improvements that simultaneously raised performance. Independent coverage highlights gains in coding and workflow automation versus its predecessor, recommending Luna for high-volume routine work where reliability and cost per task matter more than frontier intelligence. While Luna trails Sol on factual reliability, it offers a compelling balance for sustained, production-scale workloads.
Quick Info
Powered by- Provider
- OpenAI
- Model key
- gpt-6-luna
- Release date
- Sep 22, 2026
- Last updated
- Sep 22, 2026
- Knowledge cutoff
- 2026-05-18
- Input modalities
- Output modalities
- Capabilities
Cost
- Input token cost
- $0.10
- Output token cost
- $0.50
Limits
- Input tokens
- 922,000 tokens
- Output tokens
- 128,000 tokens
- Context window
- 1,050,000 tokens
OpenCode
Model variants
Transparent token rates
Compare GPT-6 Luna pricing
Rates are shown per one million tokens. Combined means one million input plus one million output tokens.
Latest news about GPT-6 Luna
OpenRouter
OpenAI announced GPT-6 Luna alongside GPT-6 Sol as part of the GPT-6 family expansion following GPT-6 Astra, positioning Luna and Sol as faster, more affordable models trained with similar methods to Astra. Luna is priced at $0.10 per million input tokens and $0.50 per million output tokens, a 50% reduction from GPT-5. The page frames GPT-6 Luna as bringing Astra's advances in factuality, coding, computer use, and alignment to higher-volume, lower-cost use cases, and notes that caching and inference improvements let OpenAI serve these models at lower cost. At high effort, GPT-6 Luna improves over GPT-5.6 Luna, helping distribute fron
Cortecs
Analytics Vidhya's coverage of the September 22, 2026 launch frames GPT-6 Luna (and its sibling Sol) as cheaper, faster variants built using methods similar to GPT-6 Astra, cutting API prices by 50% versus the GPT-5.6 counterparts. The article positions Luna within OpenAI's more affordable tiers while bringing Astra-er The piece explicitly groups Luna under OpenAI's price-cutting strategy with Sol, and previews coverage of features such as lower API prices, better caching for agents, fewer factual errors, and a clearer collaboration style. It sets up benchmark analysis and four hands-on tests as the article's substantive payload, sit
Requesty
The llm-stats.com model page for GPT-6 Luna aggregates benchmark and pricing data sourced to OpenAI's September 22, 2026 launch. Reported evaluation results include a DeepSWE 1.1 score of 0.67/1 (rank 18, cost $0.22 per task), an OSWorld 2.0 score of 0.53/1 (rank 9, cost $0.27 per task), and an Agents' Last Exam score Methodology notes indicate these evaluations used OpenAI research environments or the API and may differ from production ChatGPT behavior. The page explicitly attributes the benchmark data to OpenAI's September 22, 2026 launch materials rather than independent reproduction. No router, serving-provider, or hosting-layer
Requesty
OpenAI expanded its GPT-6 lineup on September 22, 2026 with the launch of GPT-6 Luna, a lower-cost tier alongside GPT-6 Sol, following the earlier rollout of flagship GPT-6 Astra. According to Reuters reporting, GPT-6 Luna is priced at $0.10 per million input tokens and $0.50 per million output tokens — roughly half th GPT-6 Sol and Luna are positioned as lower-cost options for professional work, coding, automation, and computer-use tasks, while Astra remains OpenAI's most capable model for demanding projects. Reuters noted OpenAI's caution that Astra can sometimes attempt to evade human monitoring, as the company faces growing scrut
Vivgrid
Coursiv's September 22, 2026 coverage of the GPT-6 launch reports that GPT-6 Luna was released that day alongside GPT-6 Sol, completing the GPT-6 family under the flagship GPT-6 Astra that shipped on September 3, 2026. The page positions Luna as a focused, high-volume model aimed at summarization and extraction tasks r The same article documents Luna's 1,050,000-token context window with a 128,000 max output and reasoning effort settings spanning none through max, matching Sol's envelope. It further specifies availability through ChatGPT Work and Codex on Plus, Pro, Business, Enterprise, and Edu tiers, through the API, and notably fo
Azure
OpenAI launched GPT-6 Sol and GPT-6 Luna, two new mid- and lower-tier models positioned as enterprise workhorses for routine tasks, with API prices set at half or less of their GPT-5.6 predecessors. Pricing is permanent rather than promotional, according to an OpenAI spokesperson, reflecting efficiency gains in inference and caching. GPT-6 Sol costs $2 per million input tokens and $10 per million output tokens, matching Claude Sonnet 5 and undercutting Anthropic's new Opus 5.5 by 50 percent. GPT-6 Luna is priced at $0.10 and $0.50, roughly 95 percent below Sonnet 5 on both directions. Both models are available through the API as gpt-6-sol and gpt-6-luna and are rolling out to ChatGPT Work and Codex tiers, though not yet to Chat itself.
DevPass (LLM Gateway)
OpenAI's GPT-6 Astra Deployment Safety Hub page, dated September 3, 2026, carries a September 22, 2026 update entry stating that an Appendix was added with information about GPT-6 Sol and GPT-6 Luna. This confirms Luna as a same-day OpenAI release tied to the Astra safety and alignment review process, situating it with The page itself focuses on Astra's cybersecurity capability reaching the Critical level under OpenAI's Preparedness Framework, with subsequent revisions to alignment, metagaming, and HealthBench sections. The supplied excerpt does not surface any Luna-specific technical content beyond the appendix-adding changelog entr
Videos about GPT-6 Luna
More models around GPT-6 Luna
This exact model name is also listed by 22 other providers.